Earth system models of intermediate complexity: closing the gap in the spectrum of climate system models

Page created by Shane Lynch
 
CONTINUE READING
Climate Dynamics (2002) 18: 579±586
DOI 10.1007/s00382-001-0200-1

M. Claussen á L. A. Mysak á A. J. Weaver á M. Cruci®x
T. Fichefet á M.-F. Loutre á S. L. Weber á J. Alcamo
V. A. Alexeev á A. Berger á R. Calov á A. Ganopolski
H. Goosse á G. Lohmann á F. Lunkeit á I. I. Mokhov
V. Petoukhov á P. Stone á Z. Wang

Earth system models of intermediate complexity: closing
the gap in the spectrum of climate system models

Received: 14 December 2000 / Accepted: 24 August 2001 / Published online: 18 December 2001
Ó Springer-Verlag 2001

Abstract We propose a new perspective on the hierar-              tion of several model types, from conceptual to com-
chy of climate models which goes beyond the ``classical''         prehensive, is presented in a new spectrum of climate
climate modeling pyramid that is restricted mainly to             system models. In particular, the location of the Earth
atmospheric processes. Most notably, we introduce a               system Models of Intermediate Complexity (EMICs) in
new indicator, called ``integration'', which characterizes        this spectrum is discussed in some detail and examples
the number of interacting components of the climate               are given, which indicate that there is currently a broad
system being explicitly described in a model. The loca-           range of EMICs in use. In some EMICs, the number of
                                                                  processes and/or the detail of description is reduced for
                                                                  the sake of simulating the feedbacks between as many
M. Claussen (&) á R. Calov á A. Ganopolski á V. Petoukhov
Potsdam-Institut fuÈr Klimafolgenforschung, Germany               components of the climate system as feasible. Others,
E-mail: claussen@pik-potsdam.de                                   with a lesser degree of interaction, or ``integration'', are
M. Claussen                                                       used for long-term ensemble simulations to study spe-
Institut fuÈr Meteorologie, FU-Berlin, Germany                    ci®c aspects of climate variability. EMICs appear to be
L. A. Mysak á Z. Wang
                                                                  closer to comprehensive coupled models of atmospheric
Department of Atmospheric and Oceanic Sciences,                   and oceanic circulation (CGCMs) than to ``conceptual''
McGill University, Montreal, Canada                               or ``box'' models. We advocate that EMICs be consid-
A. J. Weaver                                                      ered as complementary to CGCMs and conceptual
School of Earth and Ocean Sciences,                               models, because we believe that there is an advantage of
University of Victoria, Canada                                    having a spectrum of climate system models which are
M. Cruci®x á T. Fichefet á M.-F. Loutre á A. Berger á H. Goosse   designed to tackle speci®c aspects of climate and which
Institut d'Astronomie et de GeÂophysique Georges LemaõÃ tre,      together provide the proper tool for climate system
UniversiteÂCatholique de Louvain, Belgium                         modeling.
S. L. Weber
Royal Netherland Meteorological Institute,
The Netherlands
                                                                  1 Introduction
J. Alcamo
Center for Environmental Systems Research,
University of Kassel, Germany                                     Following the traditional concept of Hann (1908), cli-
                                                                  mate has been considered as the sum of all meteoro-
V. A. Alexeev
Danish Center for Earth System Science,                           logical phenomena which characterize the mean state of
Copenhagen, Denmark                                               the atmosphere at any point of the Earth's surface. This
G. Lohmann á F. Lunkeit                                           classical de®nition has proven to be useful for clima-
Meteorologisches Institut, UniversitaÈt Hamburg, Germany          tology, the descriptive view of climate. However, for
G. Lohmann
                                                                  understanding climate dynamics, i.e., the processes
Geoscience Department, UniversitaÈt Bremen, Germany               which govern the mean state of the atmosphere, the
                                                                  classical de®nition appears to be too restrictive since the
I. I. Mokhov á V. Petoukhov
A.M. Obukhov Institute of Atmospheric Physics,                    mean state of the atmosphere is a€ected by more than
Russian Academy of Sciences, Russia                               just atmospheric phenomena. In modern text books,
P. Stone
                                                                  therefore, climate is described in terms of state and
Department of Earth, Atmosphere and Planetary Sciences,           ensemble statistics of the climate system (e.g., Peixoto
Massachusetts Institute of Technology, USA                        and Oort 1992). The climate system, according to the
580                                                                     Claussen et al.: Earth system models of intermediate complexity

modern concept, consists of the abiotic world, the geo-               ``complexity'' is used here with respect to the detail of description
sphere, which is sometimes called the physical climate                and number of processes included explicitly, see later, but not in a
                                                                      mathematical sense. Also simple models can exhibit complex
system, and the living world, called the biosphere. The               behavior.) These models are simple mechanistic models which are
geosphere is further subdivided into open systems,                    designed to demonstrate the plausibility of processes. For example,
namely, the atmosphere, the hydrosphere (mainly the                   Paillard (1998) reproduced the long-term climate variations during
oceans, but also rivers), the cryosphere (inland ice, sea             the last one million years, i.e., the variations between rather short
                                                                      interglacial and longer glacials, by exploring the consequences of
ice, permafrost and snow cover), the pedosphere (the                  simple hypotheses. He assumed that long-term climate changes are
soils), and the lithosphere (the Earth's crust and the more           triggered by summer insolation at northern high latitudes and that
¯exible upper Earth's mantle). Kraus (2000) even                      there are multiple states in the natural Earth system. The system
included human activities as part of the biosphere and,               was allowed to switch from one state to another if changes in
thus, the climate system. However, this very far-reaching             insolation exceed some ad hoc de®ned thresholds.
                                                                          Di€erentiation between conceptual and comprehensive models
view creates problems as human activities can hardly be               is certainly not new. This concept is also re¯ected by Saltzman's
described by using the thermodynamic approach.                        (1985, 1988) classi®cation of inductive and quasi-deductive models.
Therefore, Schellnhuber (1999) and Alcamo (1994)                      Inductive models, according to Saltzman, are formulated based on
suggested the term Earth system to encompass the                      a gross understanding of the feedbacks that are likely to be
                                                                      involved. The system of equations, generally restricted to a very
anthroposphere, i.e., human activities, and the natural               few, are designed to be capable of generating the known climatic
Earth system, or synonymously, the ecosphere or the                   variations, or as many lines of observational evidence as possible.
climate system (Claussen 2000).                                       The inductive approach is, to cite Saltzman (1985), ``bound to be
   The modern concept of climate is rarely re¯ected in                looked upon as nothing more that curve ®tting, a charge that is
                                                                      fundamentally dicult to refute''. Essentially, the predictive value
discussions of climate models. For example, coupled                   of conceptual models is rather limited. Inductive models are the
general circulation models of the atmosphere and the                  opposite of comprehensive, quasi-deductive models which are, with
ocean (CGCMs) are considered as ``the most complete                   respect to their main components, derived from ®rst principles of
type of climate models currently available'' (Henderson-              hydrodynamics. Quasi-deductive models include, however, many
Sellers and McGue 1987). Recently, Grassl (2000)                     inductive components which are implicitly hidden in the paramet-
                                                                      rization of subgrid-scale processes.
stated that CGCMs will become a basis for Earth system                    The IPCC (Intergovernmental Panel on Climate Change;
models that describe the feedbacks of societies to climate            Technical Paper by Harvey et al., see: Houghton et al. 1997; see
anomaly predictions. Grassl (2000) further suggested                  also Chs. 1 and 8 of the IPCC Third Assessment Report, Watson
that parametrization of vegetation and other processes                et al. 2001) followed the terminology of simple and comprehensive
                                                                      models. The former include box models which are tuned to mimic
and boundary conditions in CGCMs has to be                            the sensitivity of comprehensive models. Simple models mainly
improved. This view potentially underestimates the role               serve to extend and interpolate, by physical reasoning, the results of
of vegetation dynamics and biogeochemical cycles in                   comprehensive models to a much larger number of di€erent
a€ecting the climate system and overestimates the                     scenarios associated with changes of boundary conditions.
importance of high spatial resolution and comprehen-
siveness.
   Here we argue that the description of the natural                  3 EMICs and the spectrum of climate system models
Earth system, or the climate system, should rely not only
on CGCMs, but on a spectrum of climate system                         To bridge the gap between conceptual, inductive, simple
models. The proposed spectrum explicitly acknowledges                 and comprehensive, quasi-deductive models, Earth sys-
the degree of ``integration'' (interaction) of various                tem models of intermediate complexity (EMICs) have
components of the climate system, and it leads to the                 been proposed. EMICs are designed to describe the
concept of models of intermediate complexity.                         natural Earth system excluding the interaction of
                                                                      humans and nature; humans appear as some external
                                                                      driving force. Hence a more appropriate acronym would
2 Models of the natural Earth system                                  be NEMICs (Natural Earth system Models of Inter-
                                                                      mediate Complexity) instead of EMICs. However, the
Marked progress has been achieved during the past decades in          latter acronym has now been widely used.
modeling the separate elements of the geosphere (e.g., Grassl 2000)      EMICs can be characterized in the following way.
and the biosphere (e.g., Cramer et al. 2000). This stimulated         EMICs include most of the processes described in
attempts to put all separate pieces together, ®rst in the form of
comprehensive coupled models of atmospheric and oceanic circu-        comprehensive models, albeit in a more reduced, i.e., a
lation, the CGCMs mentioned, and eventually in the form of cli-       more parametrized form. They explicitly simulate the
mate system models (CSMs) which include also biological and           interactions among several components of the natural
geochemical processes (e.g., Foley et al. 1998; Cox et al. 2000).     Earth system, mostly including biogeochemical cycles.
Comprehensive models of global atmospheric and oceanic
circulation describe many details of the ¯ow pattern, such as in-     On the other hand, EMICs are simple enough to allow
dividual weather systems and regional currents in the ocean. The      for long-term climate simulations over several thousands
major limitation in the application of these models to long-term      of years or even glacial cycles (with a period of some
climate studies arises from their high computational cost. Even       100000 years, e.g., GalleÂe et al. 1991), although not all
using the most powerful computers, only a very limited number of
multi-decadal experiments can be performed with such models.          are used for this purpose. Similar to comprehensive
    At the other end of the spectrum of complexity of natural Earth   models, but in contrast to conceptual models, the
system models are the conceptual or tutorial models. (The term        degrees of freedom of an EMIC exceed the number of
Claussen et al.: Earth system models of intermediate complexity                                                            581

adjustable parameters by several orders of magnitude.             their book, McGue and Henderson-Sellers (1997) still
EMICs are mainly quasi-deductive models, rather than              use the concept of the climate modeling pyramid, but
inductive models, although some of the components of              they extend the pyramid to include atmospheric chem-
an EMIC could belong the latter class.                            istry as a fourth vertex. Hence this pyramid re¯ects some
   Pictorially, we may de®ne an EMIC in terms of the              aspects of integration, albeit only with respect to atmo-
components of a three-dimensional vector (Claussen                spheric and near-surface oceanic processes. The classical
2000): integration, i.e., the number of interacting com-          pyramid could be extended to include integration of
ponents of the natural Earth system being explicitly              processes within all components of the climate system.
described in the model (hence the term integration is             This would lead to a multi-pod pyramid. To visualize
used here in the sense of integrated modeling rather than         integration with climate impact, or climate assessment,
in its original mathematical meaning), the number of              McGue and Henderson-Sellers (1997) suggest adding a
processes explicitly simulated, and the detail of descrip-        second pyramid, upside down on top of the climate
tion (See Fig. 1).                                                modeling pyramid, producing an ``hour glass''. Hence,
   Figure 1 provides a crude sketch of the location of            following this example, one could add a number of
the three model classes under discussion in the three-            pyramids, describing biospheric or oceanic models, for
dimensional space de®ned already. Typically, EMICs                example, with all the pyramids merging at their apices.
have less detail of description than comprehensive                Applying this idea to include all components of the cli-
models, but much more detail than conceptual models.              mate system would lead to a bundle of pyramids. We
EMICs include fewer processes than comprehensive                  therefore ®nd the spectrum sketched in Fig. 1 more
models, but a higher number of interacting components             convenient.
and vice versa compared to conceptual models. For                     However, it is not only the design by which the cli-
clarity we have overemphasized the di€erences between             mate modeling pyramid and our spectrum di€er, but
the model classes. In practice, there may be no such large        also the concept. Drawing a picture of a pyramid with a
gaps as suggested by the ®gure between conceptual                 particular type of models at the top could provoke some
models, EMICs and comprehensive models.                           misunderstanding. For example, Shackley et al. (1998)
                                                                  question the apical position of CGMCs in the context of
                                                                  scienti®cally modeling with respect to policy-relevant
3.1 Spectrum versus pyramid                                       knowledge on future climate change. They argue that
                                                                  the development of climate change science and global
The proposed model spectrum depicted in Fig. 1                    environmental policy frameworks occurs concurrently in
explicitly includes an indicator of the modeled interac-          a mutually supportive fashion, and they interpret the
tions between di€erent components of the climate sys-             dominant position of CGCMs as a social construct. In
tem through the vertical axis. It is our view that this           their reply to Shackley et al. (1998), Henderson-Sellers
model spectrum o€ers a more satisfactorily perspective            and McGue (1999) clearly pointed out that the climate
than the earlier published climate modeling pyramid of            modeling pyramid is tutorial and that one should not
Henderson-Sellers and McGue (1987).                              confuse the rather simple descriptive hierarchy of mod-
   The position of a model in the climate modeling                els as advocacy for CGCMs being the best models. We
pyramid (Fig. 2) indicates the complexity with which              agree with Henderson-Sellers and McGue (1999), and
major processes described in atmospheric models inter-            we assume that most climate modelers would agree,
act: The higher up the pyramid, the greater the inter-            although some reviewers of our papers might take the
actions among the processes. In the second edition of

Fig. 1. Pictorial de®nition of EMICs. Adapted from Claussen       Fig. 2. The climate modeling pyramid. Adapted from Henderson-
(2000)                                                            Sellers and McGue (1987)
582                                                                 Claussen et al.: Earth system models of intermediate complexity

opposite view. To researchers outside the climate                  Switzerland (model 1) and the USA (model 7) were
modeling community, however, the problem is appar-                 included in a Table of EMICs. (model 11 was added in
ently not that obvious as Shackley et al. (1999) men-              December 2000.) The table is available to the public on
tioned in their response to Henderson-Sellers and                  internet via http://www.pik-potsdam.de/data/emic/
McGue (1999). Also for this reason, we prefer the                 table_of_emics.pdf. A brief summary of the table of
phrase ``spectrum of climate system models'' instead of            EMICs is given in Table 2. It will be updated whenever
``hierarchy of climate models'' or ``climate modeling              there are new entries, either as updates of the models
pyramid''.                                                         already included or as new contributions. Therefore,
    Finally, we believe that the picture of merging pyra-          the reader is encouraged to contact the lead author
mids of various model systems through their tops, as               of this paper if his/her model is missing in the table of
suggested by McGue and Henderson-Sellers (1997) in                EMICs.
the case of a climate modeling and a climate assessment               Here we would like to present an analysis of the table
pyramid, is not generally valid. In some cases, results of         of EMICs to provide an overview of the broad spectrum
CGCMs are used for climate impact assessment. How-                 of EMICs, and to describe their approximate location in
ever, if climate models are to be integrated into a general        the new spectrum of climate system models suggested
assessment model, then simpli®ed, or even strongly                 above in Fig. 1.
simpli®ed, climate models are used (e.g., Alcamo et al.
1996; Bruckner et al. 1999; Prinn et al. 1999).
                                                                   4.1 Scope of EMICs

4 A survey of EMICs                                                EMICs are designed for a broad spectrum of purposes.
                                                                   Some deal with quite general studies of feedbacks within
The development of EMICs in various climate research               the climate system for the past, the present and future
centers around the world started roughly a decade ago,             scenarios on time scales of 102 to 105 years (models 2, 6,
and today there is an active group of EMIC modelers                8). Some focus more on atmosphere-ocean-vegetation
who meet once or twice a year to exchange notes on                 dynamics in the mid and high latitudes (models 3, 4) or
selected results from the models and to plan intercom-             globally (model 9) on time scales ranging from synoptic
parison simulation experiments. This series of meetings            to millennial, and others focus on the role of the large-
started at the Potsdam Institute of Climate Impact                 scale thermohaline circulation in the climate system on
Research in June 1999.                                             time scales of 10 to 103 years (model 10) and far beyond
   At the 25th General Assembly of the European                    (model 1). Model 5 addresses the problem of decadal
Geophysical Society held at Nice in April 2000, rep-               climate variability at mid latitudes, and models 7 and 11
resentatives of about eleven EMIC modeling groups                  are speci®cally designed for simulations with respect to
described the details of the components that made up               the problem of global change. The latter models also
each EMIC. It was decided to design a template which               include socio-economic components, and therefore they
will give for each EMIC its scope, the model compo-                come closest to a complete Earth system model, i.e., a
nents and performance, and selected applications. At               model in which the anthroposphere is included in some
the end of August 2000, two EMICs from Belgium                     interactive way, rather than being included through a
(models 4, 8, see Table 1), two from Canada (models                prescribed boundary condition.
6,10), two from Germany (models 2, 9), and one each
from the Netherlands (model 3), Russia (model 5),
                                                                   4.2 Location of EMICs in the spectrum of climate
                                                                       system models
Table 1. References to EMICs
                                                                   To determine the approximate location of EMICs in the
Model            Short list of references
                                                                   spectrum of climate system models depicted above, we
1: Bern 2.5D     Stocker et al. (1992), Marchal et al. (1998)      identify the vertical coordinate integration (I) with the
2: CLIMBER-2     Petoukhov et al. (2000),                          number of interacting components of the climate system
                  Ganopolski et al. (2000)                         explicitly described in a model. All models in the table of
3: EcBilt        Opsteegh et al. (1998)                            EMICs encompass an atmospheric module and an ocean
4: EcBilt-CLIO   Goosse et al. (2000)
5: IAP RAS       Petoukhov et al. (1998), Handorf et al. (1999),   module including a sea-ice module. The atmospheric
                  Mokhov et al. (2000)                             modules are labeled as EMBM: energy and moisture
6: MPM           Wang and Mysak (2000),                            balance models, DEMBM: energy and moisture balance
                  Mysak and Wang (2000)                            models including some dynamics, SDM: statistical
7: MIT           Prinn et al. (1999), Kamenkovich et al. (2000)
8: MoBidiC       Cruci®x et al. (2000a)(2000b)                     dynamical models, QG: quasi-geostrophic models, and
9: PUMA          Fraedrich et al. (1998),                          GCM: general circulation models. Atmospheric chem-
                  Maier-Reimer et al. (1993)                       istry is included only in models 7 and 11. The horizontal
10: Uvic         Weaver et al. (2000)                              resolution of atmospheric modules is given explicitly. In
11: IMAGE 2      Alcamo (1994), Alcamo et al. (1996)
                                                                   the case of a spectral model, the truncation (e.g., T21) is
Claussen et al.: Earth system models of intermediate complexity                                                                    583

Table 2. Interactive components of the climate system being implemented into EMICs (for explanation see text)
Model     Atmosphere                            Ocean                              Biosphere        Sea ice     Inland ice

 1        EMBM, 1-D(u)                          2-D(u, z), 3 basins                Bo, BT           T
 2        SDM, 2-D(u, k)-mL                     2-D(u, z) 3 basins                 Bo, BT, BV       TD          3-D, polythermal
 3        QG, 3-D, T21, L3                      3-D, 5.6°´5.6°, L12                                 T
 4        QG, 3-D, T21-L3                       3-D, 3°´3°                         BT, BV           TD
 5        SDM, 3-D 4.5°´6°, L8                  SDM, 2-D(u, k) 4.5°´6°, L3                          T
                                                 ®xed salinity
 6        EMBM, 1-D(u), land/ocean boxes        2-D(u, z), 3 basins                                 TD          2-D(u, z), isothermal
 7        SDM, 2-D(u, z)/atmospheric            3-D, 4°´1.25° to 3.75°, L15        BT               T
           chemistry
 8        QG, 2-D(u, z)-L2                      2-D(u, z), 3 basins                Bo, BT, BV       TD          2-D(u, z), isothermal
 9        GCM, 3-D, T21, L5                     3-D, 5°´5°, L11                    Bo               TD
10        DEMBM, 2-D(u, k)                      3-D, 3.6°´1.8°, L 19                                TD          3-D, polythermal
11        DEMBM, 2-D(u, k)/atmospheric          2-D(u, z), 2 basins                Bo, BT, BV       T
           chemistry

indicated. Ln refers to the number n of vertical layers.              The processes coordinate in Fig. 1 is a measure of the
Details can be found in the original literature cited in the       number of processes described in a model. This number
table of EMICs.                                                    is hard to evaluate. One could, for example, take the
   Ocean modules can be divided into fully three-                  number of prognostic and/or diagnostic equations.
dimensional oceanic circulation models (labelled as 3-D            However, this brings about the problem of specifying a
in Table 2) and variants on the zonally-averaged                   process. Should, for example, the variation of near-
formulation. Nearly all two-dimensional ocean models               surface heat ¯uxes with atmospheric stability, commonly
labelled 2-D(u, z) represent the world ocean by three              parametrized by using a simple stability function, be
two-dimensional boxes (Paci®c, Atlantic, Indian oceans)            considered a process? In view of these quandaries, we
which are coupled zonally where meridional boundaries              consider the spatial dimension of atmospheric and oce-
do not exist. An exception is model 11 which represents            anic modules as a convenient, because easily accessible,
the world ocean with two-dimensional models of                     indicator of the number of important processes explicitly
both the Indo-Paci®c Ocean and the Atlantic Ocean                  described in a model. For example, a two-dimensional,
joined by the Antarctic Circumpolar Ocean. Sea-ice                 zonally averaged atmospheric model [indicated as 2-
modules are separated into thermodynamic models (T)                D(u, k) or 2-D(u, z) in Table 2] commonly parametrizes
and thermodynamic models which include advection                   synoptic processes in terms of some large-scale di€usion.
and/or sea-ice dynamics (TD).                                      Likewise, a zonally averaged oceanic model does not
   All models compute the mass balance of snow on the              explicitly resolve wind-driven gyres or phenomena like
continents in some way. We consider, however, a mod-               the El NinÄo-Southern Oscillation. Hence these models
ule of inland ice sheets as interactive module only if ice         treat processes of major importance in the climate sys-
¯ow dynamics is described. The dimension of the ice                tem in a completely di€erent way than three-dimen-
¯ow model is indicated. Two inland-ice modules, labeled            sional models of the atmosphere and ocean (labelled 3-D
as polythermal, explicitly simulate the thermodynamics             in Table 2). In the case of a zonally averaged oceanic
of ice sheets.                                                     module which encompasses di€erent ocean basins, we
   With respect to biospheric modules, some models                 specify the dimension as 2.5. Likewise, we allocate the
include terrestrial carbon pools, others marine and ter-           dimension 2.5 to a vertically averaged atmospheric
restrial carbon pools, thereby allowing closure of the             module, if the three-dimensional structure of the atmo-
global carbon cycle, and a third group of models also              sphere is diagnosed (indicated as ml = multi layer in
describe global vegetation dynamics. (Model 11 also                Table 2) and is used for the parametrization of clouds or
describes global land cover changes due to anthropo-               the computation of radiative transfer. Box models, for
genic in¯uences.) Therefore, we decided to specify the             comparison, are given a dimension of 0.5.
biospheric modules as BO and BT, which refer to oceanic               The coordinate detail of description in Fig. 1 could be
and terrestrial carbon dynamics, respectively, and BV,             interpreted as the detail of description of processes. In
which refers to vegetation dynamics.                               this case, the indicator is very closely related to the
   We now quantify the coordinate integration in Fig. 1            mentioned processes. Therefore we suggest that the detail
by counting the number of interactive components                   of description characterizes the degree of geographical
explicitly described in a model. This tends to overem-             details or geographical integrity a model can potentially
phasize inclusion of biospheric components. We did this            capture. To avoid a linear correlation of integration and
on purpose in order to underline the important role of a           detail of description we have chosen to estimate only the
closed carbon cycle in the natural Earth system. How-              number of grid points of the atmospheric and oceanic
ever, a slightly di€erent speci®cation will not change the         modules, but not of the other modules. In many cases,
results of our analysis qualitatively.                             the spatial resolution of vegetation models or models of
584                                                           Claussen et al.: Earth system models of intermediate complexity

marine biosphere coincides with the resolution of the
atmospheric models or ocean models, respectively.
(There are exceptions to this rule. Model 2 uses a
dynamic ice-sheet module at a much higher resolution
than the atmospheric module, and in model 11, the
biosphere is described with two orders of magnitude
more spatial units than the atmosphere and ocean
combined. In these cases internal downscaling methods
are used to bridge the gaps between spatial scales.)
   Given this de®nition, detail of description di€ers from
integration and processes. Commonly, modelers choose a
proper balance between spatial resolution, i.e., detail of
description, and the type and number of processes to be
explicitly modeled. For several purposes, however, one
could imagine a simple energy balance model to be run
on a ®ne grid. Perhaps this model draws a fairly accurate
                                                             Fig. 3 Location of various models in the spectrum of models of
picture of areally distributed heating of the near-surface   the natural Earth system. The coordinates are: I, number of
atmosphere. As energy balance models commonly                interacting components of the climate system explicitly described in
describe large-scale energy transport in the atmosphere      a model; G, order of magnitude of the number of grid cells when
by a simple model of heat di€usion they cannot realis-       counting atmosphere and ocean modules only; D, cumulative
                                                             dimension, i.e., spatial dimension of the atmospheric and of the
tically represent atmospheric processes smaller than a       oceanic modules. The numbers refer to model numbers listed in
typical mixing length. The latter is given by the typical    Table 1
horizontal extent of weather systems, i.e., of the order
of 106-107 m. Hence there is a mismatch between detail
of description and number of processes described in a        AO-GCMs (atmosphere-ocean-sea ice models), and
model, which, however, could be perfectly sensible for       CSMs (climate system models). The latter encompasses
some applications.                                           biospheric modules (e.g., Cox et al. 2000).
   Given these speci®cations, we are now able to deter-
mine the position of EMICs in the spectrum of climate
system models (see Fig. 3). It becomes apparent there is     5 Conclusion
a broad range of EMICs. Tentatively, all EMICs can be
divided into three larger groups: models 3, 4, 9 are         De®nition of climate in terms of the state and statistics
simpli®ed comprehensive models. They were derived            of the climate system components is now commonly
from fully three-dimensional models, but with a coarser      accepted, at least in the community of climate modelers.
spatial resolution than the current comprehensive            Because of the broad spectrum of typical time scales of
models and a simpli®ed parametrization package. They         the di€erent components of the climate system, simula-
are designed for studying climate variability on decadal     tion of climate system dynamics requires di€erent types
and century time scales; therefore, they have to properly    of models. The classical pyramid of climate models
describe, not just to parametrize, the mechanisms of         proposed by, e.g., Henderson-Sellers and McGue
large-scale transport. Six of the EMICs (models 1, 2, 5,     (1987, see also the 2nd edition McGue and Henderson-
6, 8, 11) conceptually di€er from simpli®ed compre-          Sellers, 1997), does not re¯ect this aspect, and it appears
hensive models. There, comprehensiveness is deliber-         to be biased towards comprehensive models of atmo-
ately sacri®ced for the sake of integration. Two models      spheric and oceanic circulation. Therefore, we have
(models 7 and 10) do not precisely ®t into these de®ned      suggested replacing the hierarchy by a spectrum of cli-
categories. Model 7 is derived from comprehensive            mate system models which explicitly includes an indi-
models of the atmosphere and ocean. However, it uses         cator, known as integration, which characterizes the
a simple, box-like geometry of the oceans, and it has a      number of interacting components of the climate system
zonally averaged atmosphere. Model 10 encompasses            being explicitly described by a model.
comprehensive ocean, sea ice, and land-ice subcompo-            The location of various climate models in the spec-
nent models, but with a simpli®ed surface energy/mois-       trum has been discussed. In particular, the location of the
ture balance atmosphere model with parametrized              Earth system models of intermediate complexity
dynamical feedbacks.                                         (EMICs) has been determined. Figure 3 reveals that
   For comparison, we included in Fig. 3 a ``typical''       there is a broad range of EMICs re¯ecting the di€erences
box model in which the atmosphere, ocean, and vege-          in scope. In some EMICs, the number of processes and
tation are represented in a number of, of the order of       the detail of description is reduced for the sake of en-
ten, boxes. According to our speci®cation of processes in    hancing integration, i.e., the simulation of feedbacks
terms of a cumulative dimension D, they are assigned         between as many components of the climate system as
D = 1. Comprehensive models are also depicted as             feasible. Others, with a lesser degree of integration, are
AGCMs (atmospheric general circulation models),              used for long-term ensemble simulations to study speci®c
Claussen et al.: Earth system models of intermediate complexity                                                                      585

aspects of climate variability. There does not seem to                Cox PM, Betts RA, Jones CD, Spall SA, Totterdell IJ (2000) Will
exist a big gap between EMICs and CGCMs. Indeed,                         carbon cycle feedbacks amplify global warming in the 21st
                                                                         century? Nature 408: 184±187
some of the more comprehensive EMICs are derived                      Cramer W, Bondeau A, Woodward FI, Prentice IC, Betts RA,
from CGCMs. On the other hand, EMICs and concep-                         Brovkin V, Cox PM, Fisher V, Foley JA, Friend AD, Kucharik
tual or simple models di€er much more. This interpre-                    C, Lomas MR, Ramankutty N, Sitch S, Smith B, White A,
tation is not just a matter of perspective from which                    Young-Molling C (2000) Global response of terrestrial ecosys-
                                                                         tem structure and function to CO2 and climate change: results
Fig. 3 is drawn. It re¯ects the notion that EMICs as well                from six dynamic global vegetation models. Global Change Biol
as CGCMs tend to preserve the geographical integrity of                  (in press)
the Earth system, which is certainly not the case in                  Cruci®x M, Tulkens P, Berger A (2000a) Modelling abrupt climatic
conceptual models. Furthermore, in simple models                         change in glacial climate. In: Seidov D, Maslin M, Haupt B
being used in the IPCC (Intergovernmental Panel on                       (eds) Oceans and rapid past and future climatic changes. North-
                                                                         south connections. AGU Monogr Series (in press)
Climate Change, Houghton et al. 1997) projections, the                Cruci®x M, Tulkens P, Loutre MF, Berger A (2000b) A reference
climate sensitivity is prescribed. Most EMICs, except for                simulation for the present-day climate with a non-¯ux corrected
model 1, compute sensitivity, just as CGCMs do.                          atmosphere-ocean-sea ice model of intermediate complexity.
   Generally, we argue that there is a clear advantage in                Institut d'Astronomie et de GeÂophysique G. Lemaitre,
                                                                         Universite Catholique de Louvain, Progr Rep 2000/1, Louvain-
having available a spectrum of climate system models.                    la-Neuve, Belgium
Most EMICs are speci®cally designed for long-term                     Foley JA, Levis S, Prentice IC, Pollard D, Thompson SL (1998)
simulations over many millennia, and some are designed                   Coupling dynamic models of climate and vegetation. Global
to simulate the interaction of as many components of                     Change Biol 4: 561±580
                                                                      Fraedrich K, Kirk E, Lunkeit F (1998) Portable university model
the climate system as possible in an ecient manner.                     of the atmosphere. DKRZ Rep 16, Hamburg, Germany
Moreover, EMICs can explore the parameter space with                  Ganopolski A, Petoukhov VK, Rahmstorf S, Brovkin V, Claussen
some completeness. Thus, they are more suitable for                      M, Eliseev A, Kubatzki C (2000) CLIMBER- 2: a climate
assessing uncertainty, which CGCMs can do to a sig-                      system model of intermediate complexity. Part II: sensitivity
ni®cantly lesser extent. On the other hand, it would not                 experiments. Clim Dyn (in press)
                                                                      GalleÂe H, van Ypersele JP, Fichefet T, Tricot C, Berger A (1991)
be sensible to apply an EMIC to studies which require                    Simulation of the last glacial cycle by a coupled sectorially
high spatial resolution. EMICs can also be used to                       averaged climate - ice-sheet model. I. The climate model,
screen the phase space of climate or the history of cli-                 J Geophys Res 96: 13 139±13 161
mate to identify interesting time slices, thereby providing           Grassl H (2000) Status and improvements of coupled general cir-
                                                                         culation models. Science 288: 1991±1997
guidance for more detailed investigations to be under-                Goosse H, Selten FM, Haarsma RJ, Opsteegh JD (2000) Decadal
taken by CGCMs. For the interpretation of model                          variability in high northern latitudes as simulated by an inter-
results, however, conceptual models appear to be very                    mediate- complexity climate model. Ann Glaciol (in press)
useful (e.g., Brovkin et al. 1998).                                   Handorf D, Petoukhov VK, Dethlo€ K, Eliseev AV, Weiheimer A,
                                                                         Mokhov II (1999) Decadal climate variability in a coupled at-
                                                                         mosphere-ocean climate model of moderate complexity.
Acknowledgements L.A. Mysak's contribution to this paper was             J Geophys Res V.104D: 27 253±27 275
prepared while visiting (Fall 2000) the Institute for Climate         Hann J (1908) Handbuch der Klimatologie. 4th edn. Engelhorns,
Research, ETH ZuÈrich, whose hospitality and support are grate-          Stuttgart
fully acknowledged. Thierry Fichefet, Hugues Goosse, and Michel       Henderson-Sellers A, McGue K (1987) A climate modelling
Cruci®x are supported by the Belgian National Fund of Scienti®c          primer. 1st ed J Wiley New York
Research. The authors wish to thank the IGBP (International           Henderson-Sellers A, McGue K (1999) Concepts of good science
Geosphere-Biosphere Programme) core project oces GAIM                   in climate change modelling. Comments on Shackley S et al..,
(Global Analysis Integration and Modeling) and BAHC (Bio-                Climatic Change 38, 1998. Clim Change 42: 597±610
spheric Aspects of the Hydrological Cycles) for supporting our ®rst   Houghton JT, Meira Filho LG, Griggs DJ, Maskell K (1997) An
workshop in Potsdam, and the EGS (European Geophysical                   introduction to simple climate models used in the IPCC second
Society) for hosting the follow-up workshop.                             assessment report. IPCC Techn Pap II: 47 pp
                                                                      Kamenkovich IV, Sokolov A, Stone PH (2000) A coupled atmo-
                                                                         sphere-ocean model of intermediate complexity for climate
                                                                         change studies. MIT Joint Program on the Science and policy
References                                                               of global change, Report No 60. http://mit.edu/globalchange/
                                                                         MITJPSPGC_Rpt60.pdfClim Dyn submitted.
Alcamo J (ed) (1994) IMAGE 2.0: Integerated modeling of global        Kraus H (2000) Die AtmosphaÈre der Erde. Vieweg, Braunschweig
   climate change. Spec Iss Water Air Soil Pollut 76(1/2): 1±321      Maier-Reimer E, Mikolajewicz U, Hasselmann K (1993) Mean
Alcamo J, Kreileman GJJ, Leemans R (eds) (1996) Integrated               circulation of the Hamburg LSG OGCM and its sensitivity to
   scenarios of global change: results from the IMAGE 2.1 model.         the thermohaline surface forcing. J Phys Oceanogr 23: 731±757
   Spec Iss Global Environ Change 6(4): 255±394                       Marchal O, Stocker TF, Joos F (1998b) A latitude-depth, circula-
Brovkin V, Claussen M, Petoukhov V, Ganopolski A (1998) On the           tion-biogeochemical ocean model for paleoclimate studies.
   stability of the atmosphere-vegetation system in the Sahara/          Model development and sensitivities. Tellus 50B: 290±316
   Sahel region. J Geophys Res 103 (D24): 31613±31624                 Mokhov II, Eliseev AV, Handorf D, Petoukhov VK, Dethlo€ K,
Bruckner T, Petschel-Held G, Toth FL, FuÈssel H-M, Helm C,               Weisheimer A, Khvorostyanov DV (2000) North Atlantic
   Leimbach M, Schellnhuber H-J (1999) Climate change decision-          oscillation: diagnosis and modelling of interdecadal variability
   support and the tolerable windows approach. Environ Model             and its evolution. Izv Atmos Ocean Phys V.36 (5): 555±565
   Assessment 4: 217±234                                              McGue K, Henderson-Sellers A (1997) A climate modelling
Claussen M (2000) Earth system models. In: Ehlers E, Kra€t T             primer. 2nd edn. J Wiley and Sons, Chichester
   (eds) Understanding the Earth system: compartments, processes      Mysak LA, Wang Z (2000) The McGill paleoclimate model
   and interactions. Springer, Heidelberg Berlin New York                (MPM): A new Earth system model of intermediate complexity.
586                                                                  Claussen et al.: Earth system models of intermediate complexity

    Canadian Meteorological and Oceanographic Society Bull Soc         of climate and climatic change ± Part II. Kluwer Academic
    Can MeteÂorol OceÂanogr 28: 104±109                                Publishers; Davdrecht, pp 737±754
Opsteegh JD, Haarsma RJ, Selten FM, Kattenberg A (1998) EC-         Schellnhuber HJ (1999) `Earth system' analysis and the second
    BILT: a dynamic alternative to mixed boundary conditions in        Copernican revolution. Nature 402: C19 ± C28
    ocean models. Tellus 50A: 348±367                               Shackley S, Young P, Parkinson S, Wynne B (1998) Uncertainty,
Paillard D (1998) The timing of Pleistocene glaciations from a         complexity and concepts of good science in climate change
    simple multiple-state climate model. Nature 391: 378±381           modelling: are GCMs the best tools? Clim Change 38: 159±205
Peixoto JP, Oort AH (1992) Physics of climate. American Institute   Shackley S, Young P, Parkinson S (1999) Response to A. Hen-
    of Physics, New York                                               derson-Sellers and K. McGue. Clim Change 42: 611±617
Petoukhov VK, Mokhov II, Eliseev AV, Semenov VA (1998) The          Stocker TF, Wright DG, Mysak LA (1992) A zonally averaged,
    IAP RAS global climate model. Dialogue-MSU, Moscow;                coupled ocean-atmosphere model for paleoclimate studies.
    110 pp                                                             J Clim 5: 773±797
Petoukhov VK, Ganopolski A, Brovkin V, Claussen M, Eliseev A,       Wang Z, Mysak LA (2000) A simple coupled atmosphere-ocean-
    Kubatzki C, Rahmstorf S (2000) CLIMBER- 2: a climate sys-          sea ice-land surface model for climate and paleoclimate studies.
    tem model of intermediate complexity. Part I: model descrip-       J Clim 13: 1150±1172
    tion and performance for present climate. Clim Dyn 16: 1±17     Watson R, Houghton J, Yihui D (2001) climate change 200: the
Prinn R, Jacoby H, Sokolov A, Wang C, Xiao X, Yang Z, Eckaus           scienti®c basis. Contribution of working group I to the third
    R, Stone P, Ellerman D, Melillo J, Fitzmaurice J, Kicklighter      assessment report of the IPCC. Cambridge University Press,
    D, Holian G, Liu Y (1999) Integrated global system model for       Cambridge, UK
    climate policy assessment: feedbacks and sensitivity studies.   Weaver AJ, Eby M, Wiebe EC, Bitz CM, Du€y PB, Ewen TL,
    Clim Change 41: 469±546                                            Fanning AF, Holland MM, MacFadyen A, Matthews HD,
Saltzman B (1985) Paleoclimatic modeling. In: Hecht AD (ed) Pa-        Meissner KJ, Saenko O, Schmittner A, Wang H, Yoshimori M
    leoclimate analysis and modeling. Wiley New York, pp 341±396       (2000) The UVic Earth System Climate Model: model
Saltzman B (1988) Modelling the slow climatic attractor. In:           description, climatology and applications to past, present and
    Schlesinger ME (ed) Physically-based modelling and simulation      future climates. Atmosphere-Ocean (in press)
You can also read